WLM in vPars on separate nPars using iCAP and SG!

[Terence Kumar_1]

Hi

I am planning to use WLM together with vPars and iCAP on a Serviceguard cluster. I have configured 1 nPar each on a Superdome. Each nPar has 4 active and 4 hp-owned CPUs. Therefore in total, i have 8 CPUs to use in the Superdome (these are PA8800 CPUs, dual core).

Now, i want to configure a vPar on each nPar. These vPars will be clustered together. On each vPar i want to have 1 workload group consuming 2GB and 2CPU memory as a maximum. In the event of failure of npar, i am hoping to get the surviving vPar to run BOTH workload groups, each taking as a max 2GB/2CPU. Basically, the surviving vPar should resize itself to cater for this increase.

Is this even possible??? WLM is A.03.00, vPars is A.03.02 and SG is A.11.16

rgds

[melvyn burnard]

Well it looks possible, especiialy if you are using TICAP
I would recommend first reading teh various white papers available at http://docs.hp.com/en/netsys.html

Specifically:
http://www.hp.com/products1/unix/operating/docs/vpars.resizing.with.wlm.pdf

http://www.hp.com/products1/unix/operating/docs/npars.resizing.with.wlm.pdf

http://www.hp.com/products1/unix/operating/docs/wlm.serviceguard.pdf

One thing I am unceratin of here, i swhether you are using the Npars/Vpars in teh same SuperDome, or if there are two SuperDomes involved here.
If only one, then I recommend you also read

http://docs.hp.com/en/B3936-90058/B3936-90058.pdf

[Terence Kumar_1]

Hi melvyn

The thing is, there isnt any TiCAP license. Only iCAP, and yes it is a single Superdome.

In the event of failure of npar, this would "release" 4 cpus for use in the complex right? So WLM could use just iCAP to temporarily use these CPUs in the surviving partition?

rgds

[melvyn burnard]

Well yes it would, but then there woul dbe no cpu's available for the partition that failed to use at boot time as far as I can tell.

[Florian Heigl (new acc)]

This should be possible, but You should be aware that not many customers really *use* the dynamic allocation options to this extent :)

[Terence Kumar_1]

Hi melvyn

Point noted. I think i will use smaller vpars and resize only during failover. This will prevent 0 CPU's being left when the failed node tries to boot.

Hi Florian

I know i am not the first :)

cheers